wide range capacitance to digital converter with 1fF accuracy

Hi everyone

I am looking for a CDC with the following requirements

- Cin up to 70pF. Please note : the variable portion is extremely small. Cin is the sum of a large and constant capacitor (up to 70pF) and a very small varying capacitor (typical range = 0 to 300fF)

The constant and variable capacitors cannot be physically differentiated

- accuracy (and not resolution) of 1fF

I have considered the CDC AD714X and AD715Y, also with an application note AN1585 for extending the range of Cin using the AD7745 or 46, but it seems it does not match our requirements of 1fF accuracy

Any trick to measure our very small varying capacitors with 1fF accuracy embedded into a buffer and constant much larger capacitor ?

Thank you for your lights


  • 0
    •  Analog Employees 
    on Nov 13, 2020 5:20 AM 2 months ago

    You have to find a way to negate the value of Cin, and also amplify the Excitation signal depending on the small varying capacitor. such that the charge detected by the CDC can be within the full scale range.

  • It makes sense thank you very much for your feedback

    This is exactly the role of the CAPDAC feature available with the CDC AD714X and AD715Y

    However the negative CAPDAC capacitance is limited to 20pF. Then an application note AN1585 explains how to amplify the excitation signal as you suggest and to accept thus larger offset capacitance. 

    But then the remaining issue is the accuracy which is out of our fF requirement, and all the more since we have to exploit the CAPDAC & amplifying solutions. 

    So I am interested to know if other CDC are or will be available to meet our requirements, including the accuracy. 

    Cheers and take care



  • 0
    •  Analog Employees 
    on Nov 25, 2020 1:07 PM 1 month ago in reply to BBoutaud258

    There are no CDC which will meet the 1fF requirement.

    What will help in your case is if you are able to amplify the EXC signal for the small varying capacitance by X times such that it is within the full scale input capacitance range, then the measurement accuracy at the sensor level will improve by X times assuming that there are no external noise contributing factors, because at the part level, the worst case accuracy will still remain same..